1 //===-- DWARFCompileUnit.cpp ------------------------------------*- C++ -*-===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 #include "DWARFCompileUnit.h"
11 
12 #include "Plugins/Language/ObjC/ObjCLanguage.h"
13 #include "lldb/Core/DumpDataExtractor.h"
14 #include "lldb/Core/Mangled.h"
15 #include "lldb/Core/Module.h"
16 #include "lldb/Core/Timer.h"
17 #include "lldb/Host/StringConvert.h"
18 #include "lldb/Symbol/CompileUnit.h"
19 #include "lldb/Symbol/LineTable.h"
20 #include "lldb/Symbol/ObjectFile.h"
21 #include "lldb/Utility/Stream.h"
22 #include "lldb/Utility/StreamString.h"
23 
24 #include "DWARFDIECollection.h"
25 #include "DWARFDebugAbbrev.h"
26 #include "DWARFDebugAranges.h"
27 #include "DWARFDebugInfo.h"
28 #include "DWARFFormValue.h"
29 #include "LogChannelDWARF.h"
30 #include "NameToDIE.h"
31 #include "SymbolFileDWARF.h"
32 #include "SymbolFileDWARFDebugMap.h"
33 #include "SymbolFileDWARFDwo.h"
34 
35 using namespace lldb;
36 using namespace lldb_private;
37 using namespace std;
38 
39 extern int g_verbose;
40 
41 DWARFCompileUnit::DWARFCompileUnit(SymbolFileDWARF *dwarf2Data)
42     : m_dwarf2Data(dwarf2Data), m_abbrevs(NULL), m_user_data(NULL),
43       m_die_array(), m_func_aranges_ap(), m_base_addr(0),
44       m_offset(DW_INVALID_OFFSET), m_length(0), m_version(0),
45       m_addr_size(DWARFCompileUnit::GetDefaultAddressSize()),
46       m_producer(eProducerInvalid), m_producer_version_major(0),
47       m_producer_version_minor(0), m_producer_version_update(0),
48       m_language_type(eLanguageTypeUnknown), m_is_dwarf64(false),
49       m_is_optimized(eLazyBoolCalculate), m_addr_base(0),
50       m_ranges_base(0), m_base_obj_offset(DW_INVALID_OFFSET) {}
51 
52 DWARFCompileUnit::~DWARFCompileUnit() {}
53 
54 void DWARFCompileUnit::Clear() {
55   m_offset = DW_INVALID_OFFSET;
56   m_length = 0;
57   m_version = 0;
58   m_abbrevs = NULL;
59   m_addr_size = DWARFCompileUnit::GetDefaultAddressSize();
60   m_base_addr = 0;
61   m_die_array.clear();
62   m_func_aranges_ap.reset();
63   m_user_data = NULL;
64   m_producer = eProducerInvalid;
65   m_language_type = eLanguageTypeUnknown;
66   m_is_dwarf64 = false;
67   m_is_optimized = eLazyBoolCalculate;
68   m_addr_base = 0;
69   m_base_obj_offset = DW_INVALID_OFFSET;
70 }
71 
72 bool DWARFCompileUnit::Extract(const DWARFDataExtractor &debug_info,
73                                lldb::offset_t *offset_ptr) {
74   Clear();
75 
76   m_offset = *offset_ptr;
77 
78   if (debug_info.ValidOffset(*offset_ptr)) {
79     dw_offset_t abbr_offset;
80     const DWARFDebugAbbrev *abbr = m_dwarf2Data->DebugAbbrev();
81     m_length = debug_info.GetDWARFInitialLength(offset_ptr);
82     m_is_dwarf64 = debug_info.IsDWARF64();
83     m_version = debug_info.GetU16(offset_ptr);
84     abbr_offset = debug_info.GetDWARFOffset(offset_ptr);
85     m_addr_size = debug_info.GetU8(offset_ptr);
86 
87     bool length_OK = debug_info.ValidOffset(GetNextCompileUnitOffset() - 1);
88     bool version_OK = SymbolFileDWARF::SupportedVersion(m_version);
89     bool abbr_offset_OK =
90         m_dwarf2Data->get_debug_abbrev_data().ValidOffset(abbr_offset);
91     bool addr_size_OK = ((m_addr_size == 4) || (m_addr_size == 8));
92 
93     if (length_OK && version_OK && addr_size_OK && abbr_offset_OK &&
94         abbr != NULL) {
95       m_abbrevs = abbr->GetAbbreviationDeclarationSet(abbr_offset);
96       return true;
97     }
98 
99     // reset the offset to where we tried to parse from if anything went wrong
100     *offset_ptr = m_offset;
101   }
102 
103   return false;
104 }
105 
106 void DWARFCompileUnit::ClearDIEs(bool keep_compile_unit_die) {
107   if (m_die_array.size() > 1) {
108     // std::vectors never get any smaller when resized to a smaller size,
109     // or when clear() or erase() are called, the size will report that it
110     // is smaller, but the memory allocated remains intact (call capacity()
111     // to see this). So we need to create a temporary vector and swap the
112     // contents which will cause just the internal pointers to be swapped
113     // so that when "tmp_array" goes out of scope, it will destroy the
114     // contents.
115 
116     // Save at least the compile unit DIE
117     DWARFDebugInfoEntry::collection tmp_array;
118     m_die_array.swap(tmp_array);
119     if (keep_compile_unit_die)
120       m_die_array.push_back(tmp_array.front());
121   }
122 
123   if (m_dwo_symbol_file)
124     m_dwo_symbol_file->GetCompileUnit()->ClearDIEs(keep_compile_unit_die);
125 }
126 
127 //----------------------------------------------------------------------
128 // ParseCompileUnitDIEsIfNeeded
129 //
130 // Parses a compile unit and indexes its DIEs if it hasn't already been
131 // done.
132 //----------------------------------------------------------------------
133 size_t DWARFCompileUnit::ExtractDIEsIfNeeded(bool cu_die_only) {
134   const size_t initial_die_array_size = m_die_array.size();
135   if ((cu_die_only && initial_die_array_size > 0) || initial_die_array_size > 1)
136     return 0; // Already parsed
137 
138   static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
139   Timer scoped_timer(
140       func_cat,
141       "%8.8x: DWARFCompileUnit::ExtractDIEsIfNeeded( cu_die_only = %i )",
142       m_offset, cu_die_only);
143 
144   // Set the offset to that of the first DIE and calculate the start of the
145   // next compilation unit header.
146   lldb::offset_t offset = GetFirstDIEOffset();
147   lldb::offset_t next_cu_offset = GetNextCompileUnitOffset();
148 
149   DWARFDebugInfoEntry die;
150   // Keep a flat array of the DIE for binary lookup by DIE offset
151   if (!cu_die_only) {
152     Log *log(
153         LogChannelDWARF::GetLogIfAny(DWARF_LOG_DEBUG_INFO | DWARF_LOG_LOOKUPS));
154     if (log) {
155       m_dwarf2Data->GetObjectFile()->GetModule()->LogMessageVerboseBacktrace(
156           log, "DWARFCompileUnit::ExtractDIEsIfNeeded () for compile unit at "
157                ".debug_info[0x%8.8x]",
158           GetOffset());
159     }
160   }
161 
162   uint32_t depth = 0;
163   // We are in our compile unit, parse starting at the offset
164   // we were told to parse
165   const DWARFDataExtractor &debug_info_data =
166       m_dwarf2Data->get_debug_info_data();
167   std::vector<uint32_t> die_index_stack;
168   die_index_stack.reserve(32);
169   die_index_stack.push_back(0);
170   bool prev_die_had_children = false;
171   DWARFFormValue::FixedFormSizes fixed_form_sizes =
172       DWARFFormValue::GetFixedFormSizesForAddressSize(GetAddressByteSize(),
173                                                       m_is_dwarf64);
174   while (offset < next_cu_offset &&
175          die.FastExtract(debug_info_data, this, fixed_form_sizes, &offset)) {
176     //        if (log)
177     //            log->Printf("0x%8.8x: %*.*s%s%s",
178     //                        die.GetOffset(),
179     //                        depth * 2, depth * 2, "",
180     //                        DW_TAG_value_to_name (die.Tag()),
181     //                        die.HasChildren() ? " *" : "");
182 
183     const bool null_die = die.IsNULL();
184     if (depth == 0) {
185       if (initial_die_array_size == 0)
186         AddCompileUnitDIE(die);
187       uint64_t base_addr = die.GetAttributeValueAsAddress(
188           m_dwarf2Data, this, DW_AT_low_pc, LLDB_INVALID_ADDRESS);
189       if (base_addr == LLDB_INVALID_ADDRESS)
190         base_addr = die.GetAttributeValueAsAddress(m_dwarf2Data, this,
191                                                    DW_AT_entry_pc, 0);
192       SetBaseAddress(base_addr);
193       if (cu_die_only)
194         return 1;
195     } else {
196       if (null_die) {
197         if (prev_die_had_children) {
198           // This will only happen if a DIE says is has children
199           // but all it contains is a NULL tag. Since we are removing
200           // the NULL DIEs from the list (saves up to 25% in C++ code),
201           // we need a way to let the DIE know that it actually doesn't
202           // have children.
203           if (!m_die_array.empty())
204             m_die_array.back().SetEmptyChildren(true);
205         }
206       } else {
207         die.SetParentIndex(m_die_array.size() - die_index_stack[depth - 1]);
208 
209         if (die_index_stack.back())
210           m_die_array[die_index_stack.back()].SetSiblingIndex(
211               m_die_array.size() - die_index_stack.back());
212 
213         // Only push the DIE if it isn't a NULL DIE
214         m_die_array.push_back(die);
215       }
216     }
217 
218     if (null_die) {
219       // NULL DIE.
220       if (!die_index_stack.empty())
221         die_index_stack.pop_back();
222 
223       if (depth > 0)
224         --depth;
225       if (depth == 0)
226         break; // We are done with this compile unit!
227 
228       prev_die_had_children = false;
229     } else {
230       die_index_stack.back() = m_die_array.size() - 1;
231       // Normal DIE
232       const bool die_has_children = die.HasChildren();
233       if (die_has_children) {
234         die_index_stack.push_back(0);
235         ++depth;
236       }
237       prev_die_had_children = die_has_children;
238     }
239   }
240 
241   // Give a little bit of info if we encounter corrupt DWARF (our offset
242   // should always terminate at or before the start of the next compilation
243   // unit header).
244   if (offset > next_cu_offset) {
245     m_dwarf2Data->GetObjectFile()->GetModule()->ReportWarning(
246         "DWARF compile unit extends beyond its bounds cu 0x%8.8x at "
247         "0x%8.8" PRIx64 "\n",
248         GetOffset(), offset);
249   }
250 
251   // Since std::vector objects will double their size, we really need to
252   // make a new array with the perfect size so we don't end up wasting
253   // space. So here we copy and swap to make sure we don't have any extra
254   // memory taken up.
255 
256   if (m_die_array.size() < m_die_array.capacity()) {
257     DWARFDebugInfoEntry::collection exact_size_die_array(m_die_array.begin(),
258                                                          m_die_array.end());
259     exact_size_die_array.swap(m_die_array);
260   }
261   Log *log(LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
262   if (log && log->GetVerbose()) {
263     StreamString strm;
264     Dump(&strm);
265     if (m_die_array.empty())
266       strm.Printf("error: no DIE for compile unit");
267     else
268       m_die_array[0].Dump(m_dwarf2Data, this, strm, UINT32_MAX);
269     log->PutString(strm.GetString());
270   }
271 
272   if (!m_dwo_symbol_file)
273     return m_die_array.size();
274 
275   DWARFCompileUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
276   size_t dwo_die_count = dwo_cu->ExtractDIEsIfNeeded(cu_die_only);
277   return m_die_array.size() + dwo_die_count -
278          1; // We have 2 CU die, but we want to count it only as one
279 }
280 
281 void DWARFCompileUnit::AddCompileUnitDIE(DWARFDebugInfoEntry &die) {
282   assert(m_die_array.empty() && "Compile unit DIE already added");
283   AddDIE(die);
284 
285   const DWARFDebugInfoEntry &cu_die = m_die_array.front();
286   std::unique_ptr<SymbolFileDWARFDwo> dwo_symbol_file =
287       m_dwarf2Data->GetDwoSymbolFileForCompileUnit(*this, cu_die);
288   if (!dwo_symbol_file)
289     return;
290 
291   DWARFCompileUnit *dwo_cu = dwo_symbol_file->GetCompileUnit();
292   if (!dwo_cu)
293     return; // Can't fetch the compile unit from the dwo file.
294 
295   DWARFDIE dwo_cu_die = dwo_cu->GetCompileUnitDIEOnly();
296   if (!dwo_cu_die.IsValid())
297     return; // Can't fetch the compile unit DIE from the dwo file.
298 
299   uint64_t main_dwo_id = cu_die.GetAttributeValueAsUnsigned(
300       m_dwarf2Data, this, DW_AT_GNU_dwo_id, 0);
301   uint64_t sub_dwo_id =
302       dwo_cu_die.GetAttributeValueAsUnsigned(DW_AT_GNU_dwo_id, 0);
303   if (main_dwo_id != sub_dwo_id)
304     return; // The 2 dwo ID isn't match. Don't use the dwo file as it belongs to
305             // a differectn compilation.
306 
307   m_dwo_symbol_file = std::move(dwo_symbol_file);
308 
309   dw_addr_t addr_base = cu_die.GetAttributeValueAsUnsigned(
310       m_dwarf2Data, this, DW_AT_GNU_addr_base, 0);
311   dw_addr_t ranges_base = cu_die.GetAttributeValueAsUnsigned(
312       m_dwarf2Data, this, DW_AT_GNU_ranges_base, 0);
313   dwo_cu->SetAddrBase(addr_base, ranges_base, m_offset);
314 }
315 
316 dw_offset_t DWARFCompileUnit::GetAbbrevOffset() const {
317   return m_abbrevs ? m_abbrevs->GetOffset() : DW_INVALID_OFFSET;
318 }
319 
320 bool DWARFCompileUnit::Verify(Stream *s) const {
321   const DWARFDataExtractor &debug_info = m_dwarf2Data->get_debug_info_data();
322   bool valid_offset = debug_info.ValidOffset(m_offset);
323   bool length_OK = debug_info.ValidOffset(GetNextCompileUnitOffset() - 1);
324   bool version_OK = SymbolFileDWARF::SupportedVersion(m_version);
325   bool abbr_offset_OK =
326       m_dwarf2Data->get_debug_abbrev_data().ValidOffset(GetAbbrevOffset());
327   bool addr_size_OK = ((m_addr_size == 4) || (m_addr_size == 8));
328   if (valid_offset && length_OK && version_OK && addr_size_OK &&
329       abbr_offset_OK) {
330     return true;
331   } else {
332     s->Printf("    0x%8.8x: ", m_offset);
333     DumpDataExtractor(m_dwarf2Data->get_debug_info_data(), s, m_offset,
334                       lldb::eFormatHex, 1, Size(), 32, LLDB_INVALID_ADDRESS, 0,
335                       0);
336     s->EOL();
337     if (valid_offset) {
338       if (!length_OK)
339         s->Printf("        The length (0x%8.8x) for this compile unit is too "
340                   "large for the .debug_info provided.\n",
341                   m_length);
342       if (!version_OK)
343         s->Printf("        The 16 bit compile unit header version is not "
344                   "supported.\n");
345       if (!abbr_offset_OK)
346         s->Printf("        The offset into the .debug_abbrev section (0x%8.8x) "
347                   "is not valid.\n",
348                   GetAbbrevOffset());
349       if (!addr_size_OK)
350         s->Printf("        The address size is unsupported: 0x%2.2x\n",
351                   m_addr_size);
352     } else
353       s->Printf("        The start offset of the compile unit header in the "
354                 ".debug_info is invalid.\n");
355   }
356   return false;
357 }
358 
359 void DWARFCompileUnit::Dump(Stream *s) const {
360   s->Printf("0x%8.8x: Compile Unit: length = 0x%8.8x, version = 0x%4.4x, "
361             "abbr_offset = 0x%8.8x, addr_size = 0x%2.2x (next CU at "
362             "{0x%8.8x})\n",
363             m_offset, m_length, m_version, GetAbbrevOffset(), m_addr_size,
364             GetNextCompileUnitOffset());
365 }
366 
367 static uint8_t g_default_addr_size = 4;
368 
369 uint8_t DWARFCompileUnit::GetAddressByteSize(const DWARFCompileUnit *cu) {
370   if (cu)
371     return cu->GetAddressByteSize();
372   return DWARFCompileUnit::GetDefaultAddressSize();
373 }
374 
375 bool DWARFCompileUnit::IsDWARF64(const DWARFCompileUnit *cu) {
376   if (cu)
377     return cu->IsDWARF64();
378   return false;
379 }
380 
381 uint8_t DWARFCompileUnit::GetDefaultAddressSize() {
382   return g_default_addr_size;
383 }
384 
385 void DWARFCompileUnit::SetDefaultAddressSize(uint8_t addr_size) {
386   g_default_addr_size = addr_size;
387 }
388 
389 lldb::user_id_t DWARFCompileUnit::GetID() const {
390   dw_offset_t local_id =
391       m_base_obj_offset != DW_INVALID_OFFSET ? m_base_obj_offset : m_offset;
392   if (m_dwarf2Data)
393     return DIERef(local_id, local_id).GetUID(m_dwarf2Data);
394   else
395     return local_id;
396 }
397 
398 void DWARFCompileUnit::BuildAddressRangeTable(
399     SymbolFileDWARF *dwarf2Data, DWARFDebugAranges *debug_aranges) {
400   // This function is usually called if there in no .debug_aranges section
401   // in order to produce a compile unit level set of address ranges that
402   // is accurate.
403 
404   size_t num_debug_aranges = debug_aranges->GetNumRanges();
405 
406   // First get the compile unit DIE only and check if it has a DW_AT_ranges
407   const DWARFDebugInfoEntry *die = GetCompileUnitDIEPtrOnly();
408 
409   const dw_offset_t cu_offset = GetOffset();
410   if (die) {
411     DWARFRangeList ranges;
412     const size_t num_ranges =
413         die->GetAttributeAddressRanges(dwarf2Data, this, ranges, false);
414     if (num_ranges > 0) {
415       // This compile unit has DW_AT_ranges, assume this is correct if it
416       // is present since clang no longer makes .debug_aranges by default
417       // and it emits DW_AT_ranges for DW_TAG_compile_units. GCC also does
418       // this with recent GCC builds.
419       for (size_t i = 0; i < num_ranges; ++i) {
420         const DWARFRangeList::Entry &range = ranges.GetEntryRef(i);
421         debug_aranges->AppendRange(cu_offset, range.GetRangeBase(),
422                                    range.GetRangeEnd());
423       }
424 
425       return; // We got all of our ranges from the DW_AT_ranges attribute
426     }
427   }
428   // We don't have a DW_AT_ranges attribute, so we need to parse the DWARF
429 
430   // If the DIEs weren't parsed, then we don't want all dies for all compile
431   // units
432   // to stay loaded when they weren't needed. So we can end up parsing the DWARF
433   // and then throwing them all away to keep memory usage down.
434   const bool clear_dies = ExtractDIEsIfNeeded(false) > 1;
435 
436   die = DIEPtr();
437   if (die)
438     die->BuildAddressRangeTable(dwarf2Data, this, debug_aranges);
439 
440   if (debug_aranges->GetNumRanges() == num_debug_aranges) {
441     // We got nothing from the functions, maybe we have a line tables only
442     // situation. Check the line tables and build the arange table from this.
443     SymbolContext sc;
444     sc.comp_unit = dwarf2Data->GetCompUnitForDWARFCompUnit(this);
445     if (sc.comp_unit) {
446       SymbolFileDWARFDebugMap *debug_map_sym_file =
447           m_dwarf2Data->GetDebugMapSymfile();
448       if (debug_map_sym_file == NULL) {
449         LineTable *line_table = sc.comp_unit->GetLineTable();
450 
451         if (line_table) {
452           LineTable::FileAddressRanges file_ranges;
453           const bool append = true;
454           const size_t num_ranges =
455               line_table->GetContiguousFileAddressRanges(file_ranges, append);
456           for (uint32_t idx = 0; idx < num_ranges; ++idx) {
457             const LineTable::FileAddressRanges::Entry &range =
458                 file_ranges.GetEntryRef(idx);
459             debug_aranges->AppendRange(cu_offset, range.GetRangeBase(),
460                                        range.GetRangeEnd());
461           }
462         }
463       } else
464         debug_map_sym_file->AddOSOARanges(dwarf2Data, debug_aranges);
465     }
466   }
467 
468   if (debug_aranges->GetNumRanges() == num_debug_aranges) {
469     // We got nothing from the functions, maybe we have a line tables only
470     // situation. Check the line tables and build the arange table from this.
471     SymbolContext sc;
472     sc.comp_unit = dwarf2Data->GetCompUnitForDWARFCompUnit(this);
473     if (sc.comp_unit) {
474       LineTable *line_table = sc.comp_unit->GetLineTable();
475 
476       if (line_table) {
477         LineTable::FileAddressRanges file_ranges;
478         const bool append = true;
479         const size_t num_ranges =
480             line_table->GetContiguousFileAddressRanges(file_ranges, append);
481         for (uint32_t idx = 0; idx < num_ranges; ++idx) {
482           const LineTable::FileAddressRanges::Entry &range =
483               file_ranges.GetEntryRef(idx);
484           debug_aranges->AppendRange(GetOffset(), range.GetRangeBase(),
485                                      range.GetRangeEnd());
486         }
487       }
488     }
489   }
490 
491   // Keep memory down by clearing DIEs if this generate function
492   // caused them to be parsed
493   if (clear_dies)
494     ClearDIEs(true);
495 }
496 
497 const DWARFDebugAranges &DWARFCompileUnit::GetFunctionAranges() {
498   if (m_func_aranges_ap.get() == NULL) {
499     m_func_aranges_ap.reset(new DWARFDebugAranges());
500     Log *log(LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_ARANGES));
501 
502     if (log) {
503       m_dwarf2Data->GetObjectFile()->GetModule()->LogMessage(
504           log, "DWARFCompileUnit::GetFunctionAranges() for compile unit at "
505                ".debug_info[0x%8.8x]",
506           GetOffset());
507     }
508     const DWARFDebugInfoEntry *die = DIEPtr();
509     if (die)
510       die->BuildFunctionAddressRangeTable(m_dwarf2Data, this,
511                                           m_func_aranges_ap.get());
512 
513     if (m_dwo_symbol_file) {
514       DWARFCompileUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
515       const DWARFDebugInfoEntry *dwo_die = dwo_cu->DIEPtr();
516       if (dwo_die)
517         dwo_die->BuildFunctionAddressRangeTable(m_dwo_symbol_file.get(), dwo_cu,
518                                                 m_func_aranges_ap.get());
519     }
520 
521     const bool minimize = false;
522     m_func_aranges_ap->Sort(minimize);
523   }
524   return *m_func_aranges_ap.get();
525 }
526 
527 DWARFDIE
528 DWARFCompileUnit::LookupAddress(const dw_addr_t address) {
529   if (DIE()) {
530     const DWARFDebugAranges &func_aranges = GetFunctionAranges();
531 
532     // Re-check the aranges auto pointer contents in case it was created above
533     if (!func_aranges.IsEmpty())
534       return GetDIE(func_aranges.FindAddress(address));
535   }
536   return DWARFDIE();
537 }
538 
539 //----------------------------------------------------------------------
540 // Compare function DWARFDebugAranges::Range structures
541 //----------------------------------------------------------------------
542 static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
543                              const dw_offset_t die_offset) {
544   return die.GetOffset() < die_offset;
545 }
546 
547 //----------------------------------------------------------------------
548 // GetDIE()
549 //
550 // Get the DIE (Debug Information Entry) with the specified offset by
551 // first checking if the DIE is contained within this compile unit and
552 // grabbing the DIE from this compile unit. Otherwise we grab the DIE
553 // from the DWARF file.
554 //----------------------------------------------------------------------
555 DWARFDIE
556 DWARFCompileUnit::GetDIE(dw_offset_t die_offset) {
557   if (die_offset != DW_INVALID_OFFSET) {
558     if (m_dwo_symbol_file)
559       return m_dwo_symbol_file->GetCompileUnit()->GetDIE(die_offset);
560 
561     if (ContainsDIEOffset(die_offset)) {
562       ExtractDIEsIfNeeded(false);
563       DWARFDebugInfoEntry::iterator end = m_die_array.end();
564       DWARFDebugInfoEntry::iterator pos =
565           lower_bound(m_die_array.begin(), end, die_offset, CompareDIEOffset);
566       if (pos != end) {
567         if (die_offset == (*pos).GetOffset())
568           return DWARFDIE(this, &(*pos));
569       }
570     } else {
571       // Don't specify the compile unit offset as we don't know it because the
572       // DIE belongs to
573       // a different compile unit in the same symbol file.
574       return m_dwarf2Data->DebugInfo()->GetDIEForDIEOffset(die_offset);
575     }
576   }
577   return DWARFDIE(); // Not found
578 }
579 
580 size_t DWARFCompileUnit::AppendDIEsWithTag(const dw_tag_t tag,
581                                            DWARFDIECollection &dies,
582                                            uint32_t depth) const {
583   size_t old_size = dies.Size();
584   DWARFDebugInfoEntry::const_iterator pos;
585   DWARFDebugInfoEntry::const_iterator end = m_die_array.end();
586   for (pos = m_die_array.begin(); pos != end; ++pos) {
587     if (pos->Tag() == tag)
588       dies.Append(DWARFDIE(this, &(*pos)));
589   }
590 
591   // Return the number of DIEs added to the collection
592   return dies.Size() - old_size;
593 }
594 
595 // void
596 // DWARFCompileUnit::AddGlobalDIEByIndex (uint32_t die_idx)
597 //{
598 //    m_global_die_indexes.push_back (die_idx);
599 //}
600 //
601 //
602 // void
603 // DWARFCompileUnit::AddGlobal (const DWARFDebugInfoEntry* die)
604 //{
605 //    // Indexes to all file level global and static variables
606 //    m_global_die_indexes;
607 //
608 //    if (m_die_array.empty())
609 //        return;
610 //
611 //    const DWARFDebugInfoEntry* first_die = &m_die_array[0];
612 //    const DWARFDebugInfoEntry* end = first_die + m_die_array.size();
613 //    if (first_die <= die && die < end)
614 //        m_global_die_indexes.push_back (die - first_die);
615 //}
616 
617 void DWARFCompileUnit::Index(NameToDIE &func_basenames,
618                              NameToDIE &func_fullnames, NameToDIE &func_methods,
619                              NameToDIE &func_selectors,
620                              NameToDIE &objc_class_selectors,
621                              NameToDIE &globals, NameToDIE &types,
622                              NameToDIE &namespaces) {
623   Log *log(LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
624 
625   if (log) {
626     m_dwarf2Data->GetObjectFile()->GetModule()->LogMessage(
627         log,
628         "DWARFCompileUnit::Index() for compile unit at .debug_info[0x%8.8x]",
629         GetOffset());
630   }
631 
632   const LanguageType cu_language = GetLanguageType();
633   DWARFFormValue::FixedFormSizes fixed_form_sizes =
634       DWARFFormValue::GetFixedFormSizesForAddressSize(GetAddressByteSize(),
635                                                       m_is_dwarf64);
636 
637   IndexPrivate(this, cu_language, fixed_form_sizes, GetOffset(), func_basenames,
638                func_fullnames, func_methods, func_selectors,
639                objc_class_selectors, globals, types, namespaces);
640 
641   SymbolFileDWARFDwo *dwo_symbol_file = GetDwoSymbolFile();
642   if (dwo_symbol_file) {
643     IndexPrivate(dwo_symbol_file->GetCompileUnit(), cu_language,
644                  fixed_form_sizes, GetOffset(), func_basenames, func_fullnames,
645                  func_methods, func_selectors, objc_class_selectors, globals,
646                  types, namespaces);
647   }
648 }
649 
650 void DWARFCompileUnit::IndexPrivate(
651     DWARFCompileUnit *dwarf_cu, const LanguageType cu_language,
652     const DWARFFormValue::FixedFormSizes &fixed_form_sizes,
653     const dw_offset_t cu_offset, NameToDIE &func_basenames,
654     NameToDIE &func_fullnames, NameToDIE &func_methods,
655     NameToDIE &func_selectors, NameToDIE &objc_class_selectors,
656     NameToDIE &globals, NameToDIE &types, NameToDIE &namespaces) {
657   DWARFDebugInfoEntry::const_iterator pos;
658   DWARFDebugInfoEntry::const_iterator begin = dwarf_cu->m_die_array.begin();
659   DWARFDebugInfoEntry::const_iterator end = dwarf_cu->m_die_array.end();
660   for (pos = begin; pos != end; ++pos) {
661     const DWARFDebugInfoEntry &die = *pos;
662 
663     const dw_tag_t tag = die.Tag();
664 
665     switch (tag) {
666     case DW_TAG_array_type:
667     case DW_TAG_base_type:
668     case DW_TAG_class_type:
669     case DW_TAG_constant:
670     case DW_TAG_enumeration_type:
671     case DW_TAG_inlined_subroutine:
672     case DW_TAG_namespace:
673     case DW_TAG_string_type:
674     case DW_TAG_structure_type:
675     case DW_TAG_subprogram:
676     case DW_TAG_subroutine_type:
677     case DW_TAG_typedef:
678     case DW_TAG_union_type:
679     case DW_TAG_unspecified_type:
680     case DW_TAG_variable:
681       break;
682 
683     default:
684       continue;
685     }
686 
687     DWARFAttributes attributes;
688     const char *name = NULL;
689     const char *mangled_cstr = NULL;
690     bool is_declaration = false;
691     // bool is_artificial = false;
692     bool has_address = false;
693     bool has_location_or_const_value = false;
694     bool is_global_or_static_variable = false;
695 
696     DWARFFormValue specification_die_form;
697     const size_t num_attributes =
698         die.GetAttributes(dwarf_cu, fixed_form_sizes, attributes);
699     if (num_attributes > 0) {
700       for (uint32_t i = 0; i < num_attributes; ++i) {
701         dw_attr_t attr = attributes.AttributeAtIndex(i);
702         DWARFFormValue form_value;
703         switch (attr) {
704         case DW_AT_name:
705           if (attributes.ExtractFormValueAtIndex(i, form_value))
706             name = form_value.AsCString();
707           break;
708 
709         case DW_AT_declaration:
710           if (attributes.ExtractFormValueAtIndex(i, form_value))
711             is_declaration = form_value.Unsigned() != 0;
712           break;
713 
714         //                case DW_AT_artificial:
715         //                    if (attributes.ExtractFormValueAtIndex(i,
716         //                    form_value))
717         //                        is_artificial = form_value.Unsigned() != 0;
718         //                    break;
719 
720         case DW_AT_MIPS_linkage_name:
721         case DW_AT_linkage_name:
722           if (attributes.ExtractFormValueAtIndex(i, form_value))
723             mangled_cstr = form_value.AsCString();
724           break;
725 
726         case DW_AT_low_pc:
727         case DW_AT_high_pc:
728         case DW_AT_ranges:
729           has_address = true;
730           break;
731 
732         case DW_AT_entry_pc:
733           has_address = true;
734           break;
735 
736         case DW_AT_location:
737         case DW_AT_const_value:
738           has_location_or_const_value = true;
739           if (tag == DW_TAG_variable) {
740             const DWARFDebugInfoEntry *parent_die = die.GetParent();
741             while (parent_die != NULL) {
742               switch (parent_die->Tag()) {
743               case DW_TAG_subprogram:
744               case DW_TAG_lexical_block:
745               case DW_TAG_inlined_subroutine:
746                 // Even if this is a function level static, we don't add it. We
747                 // could theoretically
748                 // add these if we wanted to by introspecting into the
749                 // DW_AT_location and seeing
750                 // if the location describes a hard coded address, but we dont
751                 // want the performance
752                 // penalty of that right now.
753                 is_global_or_static_variable = false;
754                 //                              if
755                 //                              (attributes.ExtractFormValueAtIndex(dwarf2Data,
756                 //                              i, form_value))
757                 //                              {
758                 //                                  // If we have valid block
759                 //                                  data, then we have location
760                 //                                  expression bytes
761                 //                                  // that are fixed (not a
762                 //                                  location list).
763                 //                                  const uint8_t *block_data =
764                 //                                  form_value.BlockData();
765                 //                                  if (block_data)
766                 //                                  {
767                 //                                      uint32_t block_length =
768                 //                                      form_value.Unsigned();
769                 //                                      if (block_length == 1 +
770                 //                                      attributes.CompileUnitAtIndex(i)->GetAddressByteSize())
771                 //                                      {
772                 //                                          if (block_data[0] ==
773                 //                                          DW_OP_addr)
774                 //                                              add_die = true;
775                 //                                      }
776                 //                                  }
777                 //                              }
778                 parent_die = NULL; // Terminate the while loop.
779                 break;
780 
781               case DW_TAG_compile_unit:
782                 is_global_or_static_variable = true;
783                 parent_die = NULL; // Terminate the while loop.
784                 break;
785 
786               default:
787                 parent_die =
788                     parent_die->GetParent(); // Keep going in the while loop.
789                 break;
790               }
791             }
792           }
793           break;
794 
795         case DW_AT_specification:
796           if (attributes.ExtractFormValueAtIndex(i, form_value))
797             specification_die_form = form_value;
798           break;
799         }
800       }
801     }
802 
803     switch (tag) {
804     case DW_TAG_subprogram:
805       if (has_address) {
806         if (name) {
807           ObjCLanguage::MethodName objc_method(name, true);
808           if (objc_method.IsValid(true)) {
809             ConstString objc_class_name_with_category(
810                 objc_method.GetClassNameWithCategory());
811             ConstString objc_selector_name(objc_method.GetSelector());
812             ConstString objc_fullname_no_category_name(
813                 objc_method.GetFullNameWithoutCategory(true));
814             ConstString objc_class_name_no_category(objc_method.GetClassName());
815             func_fullnames.Insert(ConstString(name),
816                                   DIERef(cu_offset, die.GetOffset()));
817             if (objc_class_name_with_category)
818               objc_class_selectors.Insert(objc_class_name_with_category,
819                                           DIERef(cu_offset, die.GetOffset()));
820             if (objc_class_name_no_category &&
821                 objc_class_name_no_category != objc_class_name_with_category)
822               objc_class_selectors.Insert(objc_class_name_no_category,
823                                           DIERef(cu_offset, die.GetOffset()));
824             if (objc_selector_name)
825               func_selectors.Insert(objc_selector_name,
826                                     DIERef(cu_offset, die.GetOffset()));
827             if (objc_fullname_no_category_name)
828               func_fullnames.Insert(objc_fullname_no_category_name,
829                                     DIERef(cu_offset, die.GetOffset()));
830           }
831           // If we have a mangled name, then the DW_AT_name attribute
832           // is usually the method name without the class or any parameters
833           const DWARFDebugInfoEntry *parent = die.GetParent();
834           bool is_method = false;
835           if (parent) {
836             dw_tag_t parent_tag = parent->Tag();
837             if (parent_tag == DW_TAG_class_type ||
838                 parent_tag == DW_TAG_structure_type) {
839               is_method = true;
840             } else {
841               if (specification_die_form.IsValid()) {
842                 DWARFDIE specification_die =
843                     dwarf_cu->GetSymbolFileDWARF()->DebugInfo()->GetDIE(
844                         DIERef(specification_die_form));
845                 if (specification_die.GetParent().IsStructOrClass())
846                   is_method = true;
847               }
848             }
849           }
850 
851           if (is_method)
852             func_methods.Insert(ConstString(name),
853                                 DIERef(cu_offset, die.GetOffset()));
854           else
855             func_basenames.Insert(ConstString(name),
856                                   DIERef(cu_offset, die.GetOffset()));
857 
858           if (!is_method && !mangled_cstr && !objc_method.IsValid(true))
859             func_fullnames.Insert(ConstString(name),
860                                   DIERef(cu_offset, die.GetOffset()));
861         }
862         if (mangled_cstr) {
863           // Make sure our mangled name isn't the same string table entry
864           // as our name. If it starts with '_', then it is ok, else compare
865           // the string to make sure it isn't the same and we don't end up
866           // with duplicate entries
867           if (name && name != mangled_cstr &&
868               ((mangled_cstr[0] == '_') ||
869                (::strcmp(name, mangled_cstr) != 0))) {
870             Mangled mangled(ConstString(mangled_cstr), true);
871             func_fullnames.Insert(mangled.GetMangledName(),
872                                   DIERef(cu_offset, die.GetOffset()));
873             ConstString demangled = mangled.GetDemangledName(cu_language);
874             if (demangled)
875               func_fullnames.Insert(demangled,
876                                     DIERef(cu_offset, die.GetOffset()));
877           }
878         }
879       }
880       break;
881 
882     case DW_TAG_inlined_subroutine:
883       if (has_address) {
884         if (name)
885           func_basenames.Insert(ConstString(name),
886                                 DIERef(cu_offset, die.GetOffset()));
887         if (mangled_cstr) {
888           // Make sure our mangled name isn't the same string table entry
889           // as our name. If it starts with '_', then it is ok, else compare
890           // the string to make sure it isn't the same and we don't end up
891           // with duplicate entries
892           if (name && name != mangled_cstr &&
893               ((mangled_cstr[0] == '_') ||
894                (::strcmp(name, mangled_cstr) != 0))) {
895             Mangled mangled(ConstString(mangled_cstr), true);
896             func_fullnames.Insert(mangled.GetMangledName(),
897                                   DIERef(cu_offset, die.GetOffset()));
898             ConstString demangled = mangled.GetDemangledName(cu_language);
899             if (demangled)
900               func_fullnames.Insert(demangled,
901                                     DIERef(cu_offset, die.GetOffset()));
902           }
903         } else
904           func_fullnames.Insert(ConstString(name),
905                                 DIERef(cu_offset, die.GetOffset()));
906       }
907       break;
908 
909     case DW_TAG_array_type:
910     case DW_TAG_base_type:
911     case DW_TAG_class_type:
912     case DW_TAG_constant:
913     case DW_TAG_enumeration_type:
914     case DW_TAG_string_type:
915     case DW_TAG_structure_type:
916     case DW_TAG_subroutine_type:
917     case DW_TAG_typedef:
918     case DW_TAG_union_type:
919     case DW_TAG_unspecified_type:
920       if (name && !is_declaration)
921         types.Insert(ConstString(name), DIERef(cu_offset, die.GetOffset()));
922       if (mangled_cstr && !is_declaration)
923         types.Insert(ConstString(mangled_cstr),
924                      DIERef(cu_offset, die.GetOffset()));
925       break;
926 
927     case DW_TAG_namespace:
928       if (name)
929         namespaces.Insert(ConstString(name),
930                           DIERef(cu_offset, die.GetOffset()));
931       break;
932 
933     case DW_TAG_variable:
934       if (name && has_location_or_const_value && is_global_or_static_variable) {
935         globals.Insert(ConstString(name), DIERef(cu_offset, die.GetOffset()));
936         // Be sure to include variables by their mangled and demangled
937         // names if they have any since a variable can have a basename
938         // "i", a mangled named "_ZN12_GLOBAL__N_11iE" and a demangled
939         // mangled name "(anonymous namespace)::i"...
940 
941         // Make sure our mangled name isn't the same string table entry
942         // as our name. If it starts with '_', then it is ok, else compare
943         // the string to make sure it isn't the same and we don't end up
944         // with duplicate entries
945         if (mangled_cstr && name != mangled_cstr &&
946             ((mangled_cstr[0] == '_') || (::strcmp(name, mangled_cstr) != 0))) {
947           Mangled mangled(ConstString(mangled_cstr), true);
948           globals.Insert(mangled.GetMangledName(),
949                          DIERef(cu_offset, die.GetOffset()));
950           ConstString demangled = mangled.GetDemangledName(cu_language);
951           if (demangled)
952             globals.Insert(demangled, DIERef(cu_offset, die.GetOffset()));
953         }
954       }
955       break;
956 
957     default:
958       continue;
959     }
960   }
961 }
962 
963 bool DWARFCompileUnit::Supports_unnamed_objc_bitfields() {
964   if (GetProducer() == eProducerClang) {
965     const uint32_t major_version = GetProducerVersionMajor();
966     if (major_version > 425 ||
967         (major_version == 425 && GetProducerVersionUpdate() >= 13))
968       return true;
969     else
970       return false;
971   }
972   return true; // Assume all other compilers didn't have incorrect ObjC bitfield
973                // info
974 }
975 
976 bool DWARFCompileUnit::Supports_DW_AT_APPLE_objc_complete_type() {
977   if (GetProducer() == eProducerLLVMGCC)
978     return false;
979   return true;
980 }
981 
982 bool DWARFCompileUnit::DW_AT_decl_file_attributes_are_invalid() {
983   // llvm-gcc makes completely invalid decl file attributes and won't ever
984   // be fixed, so we need to know to ignore these.
985   return GetProducer() == eProducerLLVMGCC;
986 }
987 
988 void DWARFCompileUnit::ParseProducerInfo() {
989   m_producer_version_major = UINT32_MAX;
990   m_producer_version_minor = UINT32_MAX;
991   m_producer_version_update = UINT32_MAX;
992 
993   const DWARFDebugInfoEntry *die = GetCompileUnitDIEPtrOnly();
994   if (die) {
995 
996     const char *producer_cstr = die->GetAttributeValueAsString(
997         m_dwarf2Data, this, DW_AT_producer, NULL);
998     if (producer_cstr) {
999       RegularExpression llvm_gcc_regex(
1000           llvm::StringRef("^4\\.[012]\\.[01] \\(Based on Apple "
1001                           "Inc\\. build [0-9]+\\) \\(LLVM build "
1002                           "[\\.0-9]+\\)$"));
1003       if (llvm_gcc_regex.Execute(llvm::StringRef(producer_cstr))) {
1004         m_producer = eProducerLLVMGCC;
1005       } else if (strstr(producer_cstr, "clang")) {
1006         static RegularExpression g_clang_version_regex(
1007             llvm::StringRef("clang-([0-9]+)\\.([0-9]+)\\.([0-9]+)"));
1008         RegularExpression::Match regex_match(3);
1009         if (g_clang_version_regex.Execute(llvm::StringRef(producer_cstr),
1010                                           &regex_match)) {
1011           std::string str;
1012           if (regex_match.GetMatchAtIndex(producer_cstr, 1, str))
1013             m_producer_version_major =
1014                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
1015           if (regex_match.GetMatchAtIndex(producer_cstr, 2, str))
1016             m_producer_version_minor =
1017                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
1018           if (regex_match.GetMatchAtIndex(producer_cstr, 3, str))
1019             m_producer_version_update =
1020                 StringConvert::ToUInt32(str.c_str(), UINT32_MAX, 10);
1021         }
1022         m_producer = eProducerClang;
1023       } else if (strstr(producer_cstr, "GNU"))
1024         m_producer = eProducerGCC;
1025     }
1026   }
1027   if (m_producer == eProducerInvalid)
1028     m_producer = eProcucerOther;
1029 }
1030 
1031 DWARFCompileUnit::Producer DWARFCompileUnit::GetProducer() {
1032   if (m_producer == eProducerInvalid)
1033     ParseProducerInfo();
1034   return m_producer;
1035 }
1036 
1037 uint32_t DWARFCompileUnit::GetProducerVersionMajor() {
1038   if (m_producer_version_major == 0)
1039     ParseProducerInfo();
1040   return m_producer_version_major;
1041 }
1042 
1043 uint32_t DWARFCompileUnit::GetProducerVersionMinor() {
1044   if (m_producer_version_minor == 0)
1045     ParseProducerInfo();
1046   return m_producer_version_minor;
1047 }
1048 
1049 uint32_t DWARFCompileUnit::GetProducerVersionUpdate() {
1050   if (m_producer_version_update == 0)
1051     ParseProducerInfo();
1052   return m_producer_version_update;
1053 }
1054 
1055 LanguageType DWARFCompileUnit::LanguageTypeFromDWARF(uint64_t val) {
1056   // Note: user languages between lo_user and hi_user
1057   // must be handled explicitly here.
1058   switch (val) {
1059   case DW_LANG_Mips_Assembler:
1060     return eLanguageTypeMipsAssembler;
1061   case DW_LANG_GOOGLE_RenderScript:
1062     return eLanguageTypeExtRenderScript;
1063   default:
1064     return static_cast<LanguageType>(val);
1065   }
1066 }
1067 
1068 LanguageType DWARFCompileUnit::GetLanguageType() {
1069   if (m_language_type != eLanguageTypeUnknown)
1070     return m_language_type;
1071 
1072   const DWARFDebugInfoEntry *die = GetCompileUnitDIEPtrOnly();
1073   if (die)
1074     m_language_type = LanguageTypeFromDWARF(die->GetAttributeValueAsUnsigned(
1075         m_dwarf2Data, this, DW_AT_language, 0));
1076   return m_language_type;
1077 }
1078 
1079 bool DWARFCompileUnit::IsDWARF64() const { return m_is_dwarf64; }
1080 
1081 bool DWARFCompileUnit::GetIsOptimized() {
1082   if (m_is_optimized == eLazyBoolCalculate) {
1083     const DWARFDebugInfoEntry *die = GetCompileUnitDIEPtrOnly();
1084     if (die) {
1085       m_is_optimized = eLazyBoolNo;
1086       if (die->GetAttributeValueAsUnsigned(m_dwarf2Data, this,
1087                                            DW_AT_APPLE_optimized, 0) == 1) {
1088         m_is_optimized = eLazyBoolYes;
1089       }
1090     }
1091   }
1092   if (m_is_optimized == eLazyBoolYes) {
1093     return true;
1094   } else {
1095     return false;
1096   }
1097 }
1098 
1099 DWARFFormValue::FixedFormSizes DWARFCompileUnit::GetFixedFormSizes() {
1100   return DWARFFormValue::GetFixedFormSizesForAddressSize(GetAddressByteSize(),
1101                                                          IsDWARF64());
1102 }
1103 
1104 TypeSystem *DWARFCompileUnit::GetTypeSystem() {
1105   if (m_dwarf2Data)
1106     return m_dwarf2Data->GetTypeSystemForLanguage(GetLanguageType());
1107   else
1108     return nullptr;
1109 }
1110 
1111 void DWARFCompileUnit::SetUserData(void *d) {
1112   m_user_data = d;
1113   if (m_dwo_symbol_file)
1114     m_dwo_symbol_file->GetCompileUnit()->SetUserData(d);
1115 }
1116 
1117 void DWARFCompileUnit::SetAddrBase(dw_addr_t addr_base,
1118                                    dw_addr_t ranges_base,
1119                                    dw_offset_t base_obj_offset) {
1120   m_addr_base = addr_base;
1121   m_ranges_base = ranges_base;
1122   m_base_obj_offset = base_obj_offset;
1123 }
1124 
1125 lldb::ByteOrder DWARFCompileUnit::GetByteOrder() const {
1126   return m_dwarf2Data->GetObjectFile()->GetByteOrder();
1127 }
1128